The present disclosure relates to an image processing device and an image processing method. In particular, the present disclosure relates to an image processing device and an image processing method which enable a reduction in a data amount of a reference image which is referenced when performing motion compensation on an image, without degrading image quality of the image after motion compensation.
In relation to frame rate conversion performed using motion compensation, in motion vector detection processing and motion compensation processing, there is random access to images within a motion compensation range. However, there are many cases in which random access to the frame memory is not realizable due to the type of connection between the frame memory and a processing block and restrictions to the read-write data amounts.
In such cases, a buffer which maintains the image within the motion compensation range, which is read out from the frame memory, as the reference image is necessary in the processing block. The buffer capacity depends on the resolution of the reference image and the motion compensation range. Therefore, when the motion compensation range is expanded in order to realize support for fast motion, which is considered to be important as the basic performance of the frame rate conversion, the buffer capacity grows larger and the costs increase. Accordingly, there is a demand for a reduction in the buffer capacity.
In the motion vector detection process, the buffer capacity can be reduced by downscaling the reference image or compressing the dynamic range (the bit width) of the reference image. However, in the motion compensation processing, when the reference image is downscaled or the dynamic range of the reference image is compressed, the image quality of the image after motion compensation is degraded. Therefore, similarly to the case of the motion vector detection process, it is difficult to reduce the buffer capacity in the motion compensation processing.
Therefore, it is proposed that, using the characteristic that it is difficult to perceive a degradation of image quality in an image which has high speed motion, the reference images which correspond to high speed motion are downscaled, the reference images which correspond to static or slow motion are unchanged and motion compensation processing is performed thereon (for example, refer to Japanese Unexamined Patent Application Publication No. 2006-101267). According to the disclosure of Japanese Unexamined Patent Application Publication No. 2006-101267, the buffer capacity can be reduced while suppressing a degradation of the image quality of the image after motion compensation.
However, in the technology of Japanese Unexamined Patent Application Publication No. 2006-101267, there is a tradeoff between the buffer capacity and degradation of the image quality. In other words, when the downscale region of the reference image is increased, the buffer capacity decreases, however, since the motion amount corresponding to the downscale region increases, the motion amount range at which image quality degradation occurs increases. In contrast, when the downscale region of the reference image is reduced, the motion amount range at which image quality degradation occurs decreases, however, the buffer capacity increases.
In addition, it is proposed that the buffer capacity be reduced while suppressing the degradation of the image quality of the image after motion compensation by restoring the resolution using cubic convolution instead of the ordinary upscale process after performing the motion compensation processing using downscaled reference images (for example, refer to Japanese Unexamined Patent Application Publication No. 2009-278322). According to the disclosure of Japanese Unexamined Patent Application Publication No. 2009-278322, it is possible to suppress the degradation of the image quality of the image after motion compensation in comparison to a case in which the ordinary upscale process is performed, however, when compared to a case in which the reference images are not downscaled, there is degradation of the image quality.